Dissecting pollinator responses to a ubiquitous ultraviolet floral pattern in the wild

Authors


Summary

  1. Colour patterns on flowers can increase pollinator visitation and enhance foraging behaviour. Flowers uniform in colour to humans, however, can appear patterned to insects due to spatial variation in UV reflectance on petals. A UV ‘bullseye’ pattern that is common among angiosperms – UV-absorbing petal bases and UV-reflective apices – purportedly functions as a nectar guide, enhancing pollinator orientation and experimental evidence suggests that UV reflectance increases floral apparency to pollinators.
  2. We test the pollinator-attracting and pollinator-orienting functions of floral UV pattern and UV reflectance under natural conditions. Specifically, we address whether UV reflection alone, or UV pattern influences small bee and syrphid fly attraction rates (approaching, landing and foraging visits), foraging rates, and likelihood of foraging and orienting to the centre of flowers, using Argentina anserina, a species whose flowers exhibit variability in the size of the UV bullseye. We manipulated UV properties while maintaining uniformly yellow petals to create three phenotypes – uniformly UV-absorptive, uniformly UV-reflective, and inversed bullseye (reflective bases and absorptive apices) and compared insect visitation and behaviour to control flowers with the common UV bullseye phenotype.
  3. The presence of UV pattern increased attraction rates by both bees and syrphid flies relative to either fully UV reflective or absorptive flowers. However, only in the inverse array did the bullseye phenotype elicit higher foraging rates than the test flower. Neither the presence of pattern, nor the reversal of the common pattern influenced the likelihood of pollinator foraging or orientating to the flowers' centre during a visit.
  4. We provide some of the first evidence to suggest that flowers with spatial variation in UV reflectance may be more conspicuous to insects than those with petals that uniformly absorb or reflect UV, all of which are naturally occurring phenotypes. Further, we verify that the most common UV pattern in nature increases insect attraction and foraging rate relative to the inverse pattern. Results confirm a distance apparency function of the UV bullseye, but we argue for reconsideration of the notion that pollinators benefit from this ubiquitous floral motif through enhanced foraging efficiency.

Introduction

Circular ‘bullseye’ colour patterns on flowers can attract pollinators (Kulger 1930; Free 1970; Lehrer et al. 1995; Johnson & Dafni 1998; however see Manning 1956) and aid in their proximate orientation to the centre of a flower once they arrive (Manning 1956; Free 1970; Johnson & Dafni 1998; Dinkel & Lunau 2001), that is, act as nectar guides. Most insects are UV-perceptive (Briscoe & Chittka 2001) and bullseye floral patterns in the UV spectrum – invisible to the naked human eye – are pervasive among angiosperms (e.g. Horovitz & Cohen 1972; Guldberg & Atsatt 1975). A particular pattern whereby petal bases absorb UV while the apices reflect UV is found in many systems (e.g. Horovitz & Cohen 1972; Thompson et al. 1972; Guldberg & Atsatt 1975; Naruhashi & Ikeda 1999; Gronquist et al. 2001). As UV reflection is relevant to most insect visual systems, the assertion that this pattern functions as a nectar guide for pollinators has been widely held (Thompson et al. 1972; Eisner et al. 1973; Guldberg & Atsatt 1975; Utech & Kawano 1975). Despite this, and the fact that UV reflectance may be as relevant to insect visual systems as human-visible colour (Kevan et al. 2001), most studies have explored the effects of such patterns that are clearly visible to humans on pollinator visitation and behaviour. There is a need to test whether UV reflection and/or pattern on petals increases floral conspicuousness and whether pattern does indeed aid in pollinator orientation.

The UV bullseye on flowers manifests from UV-absorptive petal bases and UV-reflective petal apices. UV absorption at the central part of flowers creates a ‘gradient of centripetally increasing spectral purity’ and it is suggested this can enhance pollinator foraging efficiency (Lunau 1992). Some support for the nectar guide function of the UV bullseye comes from laboratory experiments where UV-absorptive regions associated with nectar rewards on false flowers elicited a foraging response in Apis mellifera (Daumer 1956). Further, Lunau and Wacht (1994) showed that the syrphid fly, Eristalis tenax, extended its proboscis over areas of purely green/yellow reflection, but the presence of UV reflection inhibited this behaviour, suggesting that UV-absorption may be important for the elicitation of foraging behaviour. In the field, various bee species (Xylocopa spp., Centris spp., Gaesischia exul, Megachile sp., Trigonia spp.) were observed landing on the UV-absorptive banner petals of asymmetric flowered Caesalpinia eriostachys and Parkinsonia aculeata (Fabaceae) (Jones & Buchmann 1974), and both UV and human-visible petal markings on Delphinium nelsonii flowers influenced bumblebee preference and behaviour (Waser & Price 1985). In none of these study species, however, did the flower possess the classic UV bullseye pattern of actinomorphic flowers, and there are limitations in extending generalities from laboratory studies to field conditions with regard to the function of floral UV pattern as a nectar guide. Behaviours elicited by UV reflection and absorption in laboratory-reared insects on artificial flowers may not hold for all flower-visiting taxa, and thus, the use of naturally occurring, non-naïve insects and natural flowers can shed light on the function of varying floral patterns in a natural pollination community. For example, Argentina anserina, the focus of this study, is only very rarely visited by honeybees or bumblebees (M. Koski, pers. obs.), and thus, these insects for which we have the most behavioural data are unlikely to be the most important taxa to influence evolution of floral traits in this system. While learning can affect the preference of pollinators for certain floral phenotypes (e.g. Laverty 1980), examining the behaviours of experienced individuals can provide a ‘real-world’ picture of how varying floral phenotypes affect insect behaviour and consequently, plant reproductive fitness.

More recent field experiments have shown that elimination of UV reflection from petals can decrease visitation by various bee species; Apis mellifera scutellata (Johnson & Andersson 2002; Welsford & Johnson 2012), Lipotriches spp. (Peter & Johnson 2008; Welsford & Johnson 2012), Bombus spp. (Rae & Vamosi 2012), and Patellapis sp. (Welsford & Johnson 2012). Male individuals of the bee fly, Megapalpus capensis, show preference for more complex patterns of UV reflective petal spots on Gorteria diffusa (de Jager & Ellis 2012). However, UV reflection did not influence visitation from the syrphid flies, Allograpta spp. or the bee, Hylaeus matamoko (Campbell et al. 2010). In most of these manipulative experiments, UV reflection was uniformly reduced across petals, eliminating any naturally occurring spatial variation of UV reflection. As a result, it remains unclear whether elimination of UV reflection or the elimination of pattern itself reduces the conspicuousness of flowers. Examining this distinction can provide insight into the evolution of floral UV traits. For instance, it is held that UV reflection can increase the conspicuousness of flowers; however, as the majority of flowers that reflect UV also have a bullseye pattern, there may also be an advantage to maintaining some degree of UV absorbance to achieve floral colour contrast (e.g. Lunau 1992; Lunau & Wacht 1994). A study that compares visitation between UV-patterned flowers and those that either uniformly absorb or reflect UV would help to clarify this issue. Which feature is most important for mediating pollinator visitation and orientation behaviour has yet to be determined for any species with UV pattern despite the fact that UV reflective, bullseye and uniformly UV-absorbing petals are all phenotypes that exist in nature (e.g. Rieseberg & Schilling 1985; Naruhashi & Ikeda 1999; M. Koski, pers. comm.). Further, whether the UV-absorbing base of petals, the most common UV flower pattern, is preferred by pollinators relative to the reverse pattern (e.g., UV-reflective petal bases and UV-absorbing apices) has not been tested in the field.

Different flower colour preference among taxa that pollinate the same species can explain the maintenance of flower colour variation in the human-visible spectrum (e.g. Streisfeld & Kohn 2007). While most important flower-visiting insects (Hymenoptera, Diptera, Lepidoptera) have visual acuity in the UV spectrum (Briscoe & Chittka 2001), sensitivities can vary among taxa. For example, the wavelength of peak UV sensitivity varies slightly among hymenopterans (Peitsch et al.1992), and some dipterans possess accessory UV pigments that may heighten their UV sensitivity relative to most hymenopterans (Warrant & Nilsson 2006; Briscoe & Chittka 2001). Thus, UV features on flowers may be more important to some taxa than others, and differential responses to varying intensity of UV or varying UV patterns on petals could be taxon-specific. Such variation may be important for generalist-pollinated plants with UV pattern variation. Indeed, both discrete and quantitative variation in the presence or size of the UV bullseye are known not only in A. anserina (Koski & Ashman 2013), but in other systems as well (Cruden 1972; Naruhashi & Ikeda 1999).

Here, we manipulate UV floral properties on the petals of A. anserina L. (Rosaceae), a widespread, generalist-pollinated plant whose uniformly yellow flowers have a bullseye in the UV spectrum (Koski & Ashman 2013). We use field experiments to compare attraction rates (number of approach, landing, and foraging visits per flower per hour), foraging rate (number of foraging visits per flower per hour), foraging behaviour (likelihood of foraging), and orientation behaviour (likelihood of orienting to the centre of the flower) of small bees and syrphid flies to flowers with petals that possess a UV bullseye pattern versus those with (i) no UV reflection/no pattern (Fig. 1a), (ii) full UV reflection/no pattern (Fig. 1b) and (iii) an inverted pattern of UV reflection (UV-absorption at the apex, reflection at the base; Fig. 1c). We address the following questions: (i) Does elimination of UV reflection and/or the elimination of the bullseye pattern reduce pollinator attraction or foraging rate and/or retard foraging and orientation behaviour? (ii) Is the common pattern (UV-absorptive flower centre) preferred, or would the inverse pattern also increase pollinator attraction or foraging rate, and enhance foraging and orientation behaviour? (iii) Do bees and flies respond similarly to different UV patterns? We discuss our results in the context of the potential adaptive function of the UV bullseye and consequences for species with variation in UV floral phenotypes.

Figure 1.

Flower types in (a) absorbing, (b) reflecting and (c) inverse arrays in the human-visible (VIS) and ultraviolet (UV) spectrum. Olfactory and tactile control flowers (O and T) in each array type had a UV bullseye phenotype, while test flowers (A, R, I) differed in UV pattern between array types. A UV-absorbing black standard is included in each photo.

Materials and methods

Study System

Argentina anserina (formerly Potentilla anserina) is a self-incompatible, hermaphroditic, stoloniferous herb that inhabits moist areas in Europe and North America (Rousi 1965). Its flowers are predominantly visited by small bees and syrphid flies (M. Koski, pers. obs.), but bumblebees have also been observed (Miyanishi, Eriksson & Wein 1991). While flowers appear uniformly yellow to humans, the apices of petals are UV-reflective while the bases are UV-absorptive and classified as ‘UV-green’ to bees (Gumbert, Kunze & Chittka 1999; Arnold et al. 2010). The area of floral UV absorption relative to flower area (hereafter, UV Proportion) is variable for A. anserina (0·30–0·99), and in the Great Lakes Region, the area for the current study, it ranges from 0·43 to 0·73 in the field (Koski & Ashman 2013).

Study Area and Plant Material

Pollinator observations at arrays of manipulated A. anserina flowers took place at Pymatuning Laboratory of Ecology (PLE) in Northwestern Pennsylvania (PLE; 41° 38′ 35·14″ N 80° 25′ 32·10″ W). Argentina anserina did not occur in the immediate vicinity of the arrays, and the only known species nearby the arrays with a UV bullseye was Ranunculus acris (Ranunculaceae) which grew in roadside ditches ~0·5 miles from the arrays (M. Koski, pers. comm.). One closely related, yellow-flowered species, Potentilla canadensis, was flowering in very low abundance near the arrays and its petals were completely UV absorbing (M. Koski, pers. obs.). Dominant flowering species at the site included Fragaria virginiana, and Rubus allegheniensis (both with actinomorphic, white, uniformly UV-absorbing flowers), Securigera varia, and Lotus sp. (both with zygomorphic, uniformly UV-absorbing flowers) (M. Koski, pers. obs.). Argentina anserina flowers used in artificial arrays were collected from two populations (41° 54′ 24·07″ N 80° 48′ 15·05″ W and 41° 51′ 09·57″ N 80° 33′ 31·65″ W) on the shore of Lake Erie in Northeastern Ohio, USA, and transported in a cooler to PLE. Flower size was not different between the populations (diam., mean ± SE; 15·9 ± 0·55 vs. 16·4 ± 0·82 mm; = −0·54, = 0·59), and collections from these were pooled into a stock ‘population’ from which we randomly allocated flowers to the following experiments.

Floral Manipulation

Three array types were created: UV absorbing (Fig. 1a), UV reflecting (Fig. 1b) and inverse UV bullseye (Fig. 1c). In each array, there were two flowers in each of three categories; two control groups (O and T) with the ‘wild-type’ phenotype of a UV bullseye, and a test group with a novel floral phenotype [absorbing (A), reflecting (R), or inverse (I); Fig. 1a–c]. One control flower accounted for olfactory changes (O), while the other controlled for both olfactory and tactile changes (T) made to the test flowers.

To achieve the absorbing test flower (A), we spread a mixture of Parsol MXC and Parsol 1789 sunscreens dissolved into duck preen gland fat (Marryat Real Duck Grease, Switzerland) (hereafter, sunscreen mixture) onto the upper side of petals with a small paintbrush (Johnson & Andersson 2002; Peter & Johnson 2008). To control for scent of the sunscreen mixture, the O flower received this same mixture on the underside of petals. To control for both scent and petal texture the T flower received duck fat on the top and the sunscreen mixture on the bottom of petals. Duck fat alone did not greatly alter the spectral properties of the petals when applied to their upper side (Figs 1a and S1a,b, Supporting information).

To achieve complete UV reflection on petals of the reflecting test flower (R), the upper surface of petals was painted with yellow UV-reflective paint (UV Yellow, Fish Vision UV Lure Paint) (Fig. 1b). Similar UV-reflective paints were used to manipulate eye-spot phenotypes on butterfly wings (Prudic et al. 2011). We applied paint to the underside of petals on the O flower to control for scent. On the T flower, we painted only the UV-reflective apices of upper surface of petals to control for scent and petal texture. The paint was scented (M. Koski, pers. comm.) and had only a slightly higher reflectance than naturally occurring tips of flowers (Fig. S1a,d,e).

For the inverse test flower (I), we painted the base of petals with yellow UV-reflective paint and the apex of petals with the sunscreen mixture. This effectively inverted the bullseye pattern in the UV (Fig. 1c). The O flowers received UV reflective paint and the sunscreen mixture on the underside of the petals. The T flowers received paint on the UV-reflective apices of petals and sunscreen on the UV-absorptive bases of petals (recreating the ‘wild-type’ bullseye phenotype). For all manipulations, care was taken to avoid spreading paint or sunscreen over nectaries and anthers. We measured spectral reflectance at the petal base and apex for all of the controls and test flowers with an Ocean Optics USB4000 spectrometer with a UV-NIR DH-2000-BAL deuterium/tungsten light source (see Fig. S1).

Array Set-Up and Pollinator Observations

We recorded pollinator responses to flowers in arrays from 17 May to 18 June, 2012. Circular arrays consisted of six flowers in total – two flowers of each type (O, T controls and an A, R or I test, as appropriate) arranged alternately (Fig. 2). Flowers were placed in water-filled microcentrifuge tube ‘aquapics’ attached to stakes to achieve natural flower height (Fig. 2). Observation periods of single arrays ranged from 45 min to 2 h. At half hour intervals, the flowers were rotated to a different location in the circular array. If a flower wilted before the observation period was completed, we replaced it with a fresh flower of the same treatment type, and to control for the presence of a fresh flower in one treatment group, we replaced a single random flower from each of the other treatments with a fresh flower. Two arrays were placed ~7 m apart and were observed simultaneously by different observers. Halfway through each observation period, flowers from one array were swapped with those from the other array to reduce spatial and observer bias. We alternated array types between morning (900–1200 h) and afternoon (1200–1600 h). In total, we observed 12 arrays with A test flowers (‘absorbing arrays’; 22·25 h of observation), 12 with R test flowers (‘reflecting arrays’; 21·75 h) and 10 with I test flowers (‘inverse arrays’; 19 h).

Figure 2.

An example of an array in which cut flowers were placed in microcentrifuge tubes elevated on florist sticks (top). A syrphid fly (bottom left) and solitary bee (bottom right) foraging on flowers of Argentina anserina in experimental arrays.

We recorded all insects that (i) approached, (ii) landed on, or (iii) landed on and foraged at a flower. Approach visits were scored if a pollinator hovered over, but did not make contact with the flower. Landing visits consisted of those in which pollinators landed on a petal (regardless of orientation) but did not forage for nectar or pollen. Visits were scored as foraging when insects clearly sought pollen and/or nectar. In general, bees foraged by pivoting over the gynoecium and androecium, or orienting their bodies horizontally at the base of the androecioum and moving in circles around it (e.g. Chagnon, Gingras & De Oliveira 1993), while flies positioned themselves over the gynoecium and androecium, or on the petal and probed at nectaries at the base of the petals (Fig. 2). Activity of flower visitors was high, making it difficult to determine whether a pollinator was making its first visit to the array, so we recorded all visits including successive visits by the same insect to a different flower in the array. However, if a pollinator left a given flower and quickly revisited it, then this was considered a single visit. Both first and subsequent visits by a pollinator reflect preferences of the insect for a floral phenotype, and it is not uncommon to consider both in data sets of pollinator behaviour (e.g. Schemske & Bradshaw 1999). We note that multiple visits from the same insect are not independent data points as a result of learned or inherent behavioural differences between individuals. We believe, however, that the data are representative of the choices of many insects as insect activity was very high at arrays and on natural flowers surrounding arrays (>200 insects in a 10 × 10 m area during a given array; M. Koski, pers. comm.) and experiments were conducted over a period >1 month. Thus, data are unlikely biased towards only a few individual insects.

For visits in which insects landed on flowers, we recorded orientation behaviour as (i) oriented to the centre of the flower (either by landing on the centre or landing on the edge and walking to the centre) or (ii) did not orient to the centre of the flower (e.g. walked across petal without walking to the centre). Across all array types, 98·5% of visits were by bees and syrphid flies, and these were recorded separately so that we could determine whether response to flowers differed between these broad groups. We were unable to identify visitors to a more detailed taxonomic distinction than ‘bee’ and ‘fly’ due to high rates of visitation. Members of the fly group included syrphid flies (Syrphidae) from two genera; Episyrphus and Sphaerophoria. Peak visual sensitivity in the UV spectrum is known in at least seven dipteran species, including one syrphid (Eristalis tenax; Horridge, Mimura & Tsukahara 1975). Bee visitors to arrays were from four families; Apidae (Epioloides sp., Holcopasites sp.), Andrenidae (Callopsis coloradensis, Callopsis sp., Perdita sp.), Megachilidae (Stelis sp.) and Halictidae (Lasioglossum spp.). The majority of Apidae species whose visual systems have been characterized are UV-sensitive, as are all Andrenidae and Megachilidae, but UV-sensitivity is not yet known for Halictidae species (Peitsch et al. 1992). Other rare visitors (Lepidoptera, ant, large fly, small muscid fly, ladybug, weevils) did not contact reproductive parts and were not considered further. Similar types of insects have been observed to visit flowers in natural populations in the Great Lakes area (M. Koski, pers. comm.).

For each replicate array, we characterized four responses to each floral phenotype for bees and flies separately as follows. We calculated total attraction rate to each flower type (visits/flower/hour) using all types of visits recorded for each group (approach without landing, land without foraging and foraging). This metric represents the degree to which insects are attracted, from a distance, to a given flower type. We then assessed foraging rate to each flower type (foraging visits/flower/hour) using only the visits in which insects foraged. This metric categorizes a ‘legitimate’ visit that is the best proxy for the pollination success of a flower. We then scored foraging behaviour as the proportion of total visits that led to foraging (hereafter ‘proportion foraging’). Finally, we scored orientation behaviour as the proportion of landing visits (landed and landed/foraged) that led to centring (hereafter ‘proportion centring’).

Statistical Analyses

For each array type, we analysed attraction rate and foraging rate using mixed-model anovas (SAS, Proc MIXED; sas v. 9.3, SAS Institute, Cary, NC) with flower treatment, pollinator type and flower × pollinator as fixed effects and replicate and all interactions with replicate as random effects. We used planned contrasts to compare rates between the controls O and T (which have the same visual phenotype), and between T and a given test (A, R or I) flower (different phenotypes) (e.g. Melendéz-Ackerman & Campbell 1998). When the two controls were not significantly different, it indicates that there was no effect of the paints/carriers on behaviours to the phenotypically identical flowers, and they were pooled and compared with the test flower. If O and T were different, this indicated that manipulation to the upper side of petals influenced flower-visitors, and thus, the comparison between T and the test flower offered the best evaluation of the sole effect of the flowers' visual phenotype. We present both the results of the pooled contrast and contrasts between the T flower and a given test (A, R or I). Attraction rate was ln transformed for the reflecting array, and attraction rate and foraging rate were ln + 1 transformed for the absorbing and inverse arrays to satisfy the assumption of normality.

To test the effect of flower treatment on foraging behaviour (proportion of total visits that resulted in foraging) in the absorbing and inverse arrays, we used mixed-model anovas (SAS, Proc MIXED) with flower, pollinator and their interaction as fixed effects, and replicate and all interactions with replicate as random effects. In the reflecting array, however, data were negatively skewed and could not be transformed to meet the assumption of normality, so we used a generalized linear mixed model (SAS, Proc GLIMMIX) with a binomial distribution and a Logit link function. When modelled with a binomial distribution the Generalized χ2/DF value was closer to one (1·12) than when modelled with a Poisson distribution (0·11), indicating that the binomial istribution was a better fit for the data (Schabenberger 2005).

To test the effect of floral manipulations on the orienting behaviour, we analysed proportion of centring visits using a generalized linear mixed model as described previously with a binomial distribution. For each array, the χ2/DF value was closer to one when using a binomial distribution (0·84–1·23) than when using a Poisson distribution (0·02–0·04).

Results

Absorbing Arrays

We observed 930 bee visits and 319 fly visits across the 12 replicates of absorbing arrays, and foraging behaviour was recorded for all 1249 visits. Of the 1004 visits in which pollinators landed on flowers, we scored orientation behaviour for 987 visits. Flower type significantly influenced attraction rate (Table 1a, Fig. 3a). Attraction rate to the fully absorbing (A) test flowers was, on average, ~13% lower than to control flowers (O and T) with a bullseye, which were not different from each other (O vs. T; Table 1a, Fig. 3a). This significant difference persisted in the comparison of the T with the A flower, attraction rate to the absorbing flower was ~10% lower (Table 1a, Fig. 3a). Foraging rate was influenced by flower type, but the difference was between the controls (Table 1a, Fig. 3d), and thus bullseye flowers did not elicit a higher rate of foraging than the fully absorbing flowers. There was no flower by pollinator-type interaction (Table 1a) effect on either attraction or foraging rate, indicating that the response to UV manipulation was similar between bees and flies.

Table 1. Results from linear models testing the effect of flower treatment [O, T, test (A, R, I)], pollinator type (Poll) and their interaction on attraction rate [(approaches + lands + forages)/flowers/hour)], foraging rate (forages/flowers/hour) proportion of foraging visits, and proportion of centring visits in each array type [(a) absorbing, (b) reflecting, (c) inverse]. Planned contrasts to test for differences between controls and test flowers (A, R, I) versus controls were used when the effect of flower was significant
Source of variation and flower contrastsNum. dfDen. dfAttraction rateForaging rate% Foraging visits% Centring visits
F-valueF-valueF-valueF-value
  1. P < 0·06, *< 0·05, **< 0·01, ***< 0·001.

(a) Absorbing
Flower220–226·91**4·53*1·670·26
O vs. T1221·446·89*
T vs. A1225·98*0
O + T vs. A12212·37**
Pollinator11113·84**17·69**13·03**14·53**
Flower × pollinator220–220·171·141·60·2
(b) Reflecting
Flower218–226·57**2·660·160·22
O vs. T1221·83
T vs. R1225·15*
O + T vs. R12211·56**
Pollinator1114·666·57*8·09*14·30**
Flower × pollinator218–220·61·610·780·01
(c) Inverse
Flower217–189·74**9·16**3·96*1·03
O vs. T1181·063·745·93*
T vs. I11810·24**5·48*0
O + T vs. I11818·41***14·59**
Pollinator197·16*9·05*1·4621·33**
Flower × pollinator217–182·753·070·511·71
Figure 3.

The effect of UV pattern on bee and fly attraction rate, foraging rate, likelihood of foraging (foraging proportion) and likelihood of orienting to centre (centring proportion) to olfactory and tactile control flowers (O and T) and test flowers (A, R, I) in three array types (absorbing, reflecting, inverse). Brackets indicate that the controls were not different and were therefore pooled and compared to the test flower (Table 1). least squares means ± SE from linear models using ln or ln + 1 transformed (visitation rate) or raw data (foraging proportion) are plotted, and P-values are derived from planned contrasts. In cases in which a generalized linear mixed model was used, LSmeans and SEs were back transformed and plotted for graphical purposes (h, j, k, m). Data points denote the floral phenotype where black represents UV-absorption and white represents UV-reflection. Note that scales on the X-axes vary. *< 0·05, **< 0·01, ***< 0·0001.

While bees were more likely to forage during a visit than flies (Table 1a), the likelihood of foraging visits was not different between flower types (Table 1a; Fig. 3g). That is, pollinators were equally likely to forage on a bullseye control flowers (O or T) as they were on absorbing (A) flowers. There was no flower by pollinator type effect on proportion foraging (Table 1a). Again, bees were more likely to orient to centre than flies once they landed on flowers (Table 1), but orientation behaviour was not influenced by flower type (Table 1a; Fig. 3j).

Reflecting Arrays

Across the 12 replicates of reflecting arrays we observed 830 bee visits and 484 fly visits (n = 1314). Foraging behaviour was scored for all but one visit (n = 1313) and orientation behaviour was scored for 1009 of the 1020 visits in which pollinators landed. Overall, flower type influenced attraction rate (Table 1b). For both pollinator groups uniform, UV reflection and elimination of pattern (R flower) decreased attraction rates significantly (~18%) relative to the controls (O and T) which did not elicit different attraction rates (Table 1b; Fig. 3b). A significant reduction to the fully reflective flower (~11%) still exists when comparing only the T and the R flower (Table 1b). Despite the fact that attraction rate was influenced by flower type, the rate of foraging visits was only marginally influenced (= 2·66, = 0·09, Table 1).

Flower type did not influenced the likelihood of pollinator foraging (Table 1b; Fig. 3h) or orientation of either pollinator group (Table 1b; Fig. 3k) but bees and flies differed in these behaviours (Table 1b).

Inverse Arrays

We observed 1077 bee and 372 fly visits across the ten replicate inverse arrays and foraging behaviour was scored for every visit (n = 1449). Orientation behaviour was scored for all visits in which pollinators landed on the flower (n = 1093). Floral manipulation affected attraction rate and foraging rate (Table 1c, Fig. 3c,f). The ‘normal’ bullseye pattern controls had the highest attraction rate (~16% higher than inverse) and did not differ from one another (Table 1c). Comparing attraction rates of the I to the T only, there was a 12% reduction for the inverse flower type. Foraging rates were lower in the inverse treatment when controls were grouped (~21%) and when only comparing the T and I flower (~15%) (Table 1c). Bees and flies responded similarly to the floral manipulations (Table 1c).

Flower type influenced the proportion of foraging visits but the difference was between the controls (O vs. T, Table 1c, Fig. 3i) rather than the tactile control and the test flower (T vs. I, Table 1c), indicating that the inverse UV pattern did not affect the probability of foraging. Flower type did not affect the likelihood of pollinators orienting to the centre of flowers, but bees and flies differed in this behaviour (Table 1c).

Discussion

We show experimentally that the presence of the floral UV bullseye pattern increases the conspicuousness of flowers to small bees and syrphid flies, but not their likelihood of foraging or orienting. Our findings are important contributions to the understanding of the role of UV reflection and pattern in mediating plant/pollinator interactions for three primary reasons. First, the presence of the UV bullseye did not increase the likelihood of insect foraging nor their ability to orient to the centre of flowers, calling into question its function as a nectar guide at close range. Second, in other studies, UV reflection alone is shown to increase insect visitation, but we found that an increased area of UV reflection on petals led to a decrease in insect attraction relative to flowers with patterned petals, and we attribute this to the elimination of pattern. Third, we confirm that the most common UV floral pattern – UV-absorbing petal bases and reflecting tips – was more conspicuous to bees and flies than the inverse pattern and increased the foraging rate but did not affect the likelihood of insect foraging or orienting to the flowers' centre. We suggest that this specific UV pattern functions to increase floral apparency from a distance but may not necessarily act as a proximate pollinator orientation guide as long proposed.

Is the UV Bullseye a Nectar Guide?

While many studies describe UV absorption at petal bases as a nectar guide (Thompson et al. 1972; Eisner et al. 1973; Guldberg & Atsatt 1975; Utech & Kawano 1975), we found that UV-absorbing petal bases had no effect on bee or fly orientation to floral rewards or their likelihood of foraging despite high power to detect these effects afforded by the large number of visitors observed and replication of each array type. In contrast, Jones and Buchmann (1974) observed various taxa of bees orienting to the UV-absorptive petal of two species (Caesalpinia eriostachys and Parkinsonia aculeata). Two factors differ between our study system and those of Jones and Buchmann which could contribute to the disparity between our findings: (i) Flowers of Argentina anserina are radially symmetric, and (ii) Floral rewards (pollen and nectar) are not concealed. Conversely, the species studied by Jones and Buchmann have irregular flowers with concealed nectar. If these differences are causal, then together, the two studies bolster the assertion that nectar guides are more important in irregular flowers than symmetric ones because nectaries of the former are more difficult for pollinators to locate (Manning 1956). We suggest that, in our system, pollen and/or scent cues (Lunau 2000; Pernal & Currie 2002; Ashman et al. 2005) may alone be effective orientation guides. Our study joins Kulger (1930) who found that bumblebees were equally as likely to locate rewards on flowers with and without central nectar guides. We therefore caution against assuming that UV patterns on petals function as orientation cues in all systems. However we acknowledge that handling time was not assessed in our study and others have shown that handling time can be reduced by the presence of nectar guides (Waser & Price 1983; Leonard & Papaj 2011).

Interestingly, laboratory and field studies have shown that the presence of a ‘target’ or bullseye can increase the ability of pollinators to orient to the centre of a flower or flower mimic (Manning 1956; Free 1970; Johnson & Dafni 1998; Dinkel & Lunau 2001). A behavioural explanation for why we did not find this is that experienced flower-visitors may be accustomed to landing on the centre of radially symmetric flowers regardless of the presence of a target. It is possible that the presence of the bullseye pattern influences learning behaviour in early life, but fails to function as a nectar guide for an experienced pollinator. However, Leonard and Papaj (2011) showed that linear markings on flower petals increased the ability of Bombus impatiens to discover nectar both immediately (inherently) and after experience with foraging. The artificial flowers used by Leonard and Papaj were about three times larger than the natural flowers of A. anserina however. ‘Nectar guides’ may be more likely to orient pollinators in larger-flowered systems, and studies that consider the effect of flower size on the magnitude of ‘nectar guide’ effectiveness would help to address this proposed idea.

UV Reflectance or Pattern: Which Mediates Visitation?

Recent studies show that floral UV reflectance mediates plant/pollinator interactions. Namely, elimination of UV reflectance from petals reduced visitation rate by various bees (Johnson & Andersson 2002; Rae & Vamosi 2012; Welsford & Johnson 2012) and reduced visitation due to loss of UV reflectance can reduce reproductive fitness (Peter & Johnson 2008). In the present study, elimination of UV reflectance and pattern from petals in the absorbing array did indeed reduce attraction rates (Fig 3a), but surprisingly, eliminating pattern by increasing the UV-reflective area on petals had the same effect (Fig. 3b). Thus, our results suggest that contrast in the UV spectrum on petals (UV pattern) may be more important in mediating insect attraction than UV reflection alone. Our work corroborates experiments by Hertz (1931) which suggest that more ‘broken’ patterns are preferred to less ‘broken’ patterns, and Kulger (1930) who found that the presence of nectar guides influences conspicuousness of flowers at a distance, but not insect orientation behaviour once in contact with the flower.

Examining only foraging rate (i.e. legitimate visits), the elimination of the bullseye did not reduce visitation in the absorbing array suggesting that, despite the fact that bullseye flowers were more conspicuous, they did not experience increased functional visitation. When pattern was eliminated such that the petals were uniformly UV reflecting, the bullseye flower tended to receive higher foraging rates (Fig. 3e). Interestingly, despite the nonsignificant overall effect of flower (= 0·09), pairwise comparisons show that the entirely reflecting flower experienced a marginally significant reduction in foraging rate (T vs. R, = 0·07, O and T vs. R, = 0·03; Fig. 3e). Thus, considering the results from the absorbing and reflecting arrays together, we can cautiously speculate that the presence of UV absorption (uniform or bullseye) could be effective at eliciting foraging behaviour, corroborating Daumer (1956), and Lunau & Wacht (1994). However, this assertion needs to be substantiated with data from arrays that simultaneously compare pollinator responses to UV-reflecting and UV-absorbing flower types.

Pervasiveness of the UV Bullseye Flower Pattern

The UV bullseye pattern is common among angiosperms and not restricted to any particular plant family (e.g., Asteraceae, Thompson et al. 1972; Skogin 1977; Brassicaceae, Horovitz & Cohen 1972; Rosaceae, Naruhashi & Ikeda 1999). The rate of foraging, a metric that only included the ‘legitimate’ visits in which insects made contact with reproductive parts of flowers, was only influenced by inverse flower type, in which the common bullseye pattern was reversed. Thus, our results suggest that there may be a fitness advantage for individuals with the common UV bullseye relative to the inverse pattern via increased pollinator foraging, however, phylogenetically-controlled tests are required to show that this phenotype is an example of convergent evolution in response to pollinator preference. Lunau (1992) suggests that a UV-absorptive flower centres can aid in recognition, orientation and landing abilities of bees and flies. We show that UV-absorptive flower centres may indeed be important for floral apparency as a result of either inherent preference for floral UV bullseye patterns or a familiarity with these in nature (floral constancy). A possible explanation for reduced attraction rates to the inversed bullseye flowers may derive from its lower contrast from the vegetative background. UV reflection from green vegetation is low (generally <5%; Caldwell, Robberecht & Flint 1983), so flowers with UV-reflective petal apices may be more apparent to insects. Reduced contrast from the background for the inverse flower may have resulted in a visually smaller flower than those with the common UV bullseye.

We observed behaviours of wild insects and do not know their level of experience. Given that the pollinators were not necessarily naïve, we can not address whether the behaviour we recorded was innate or learned. However, others have shown that naïve bees show a preference for patterned flower mimics as opposed to non-patterned flower mimics (Free 1970; Lehrer et al. 1995; Leonard & Papaj 2011). In the field, visitation to both mimic and real flowers is increased by the presence of pattern (bees; Hansen, Van der Niet & Johnson 2011; flies; Johnson & Dafni 1998). Pollinators in the present study may have been experienced with other flowers that possess the more common UV bullseye pattern because Ranunculus acris, which has a yellow, UV bullseye flower, did grow in the area. The dominant flowering plants within the immediate vicinity of arrays, however, did not possess the UV bullseye pattern (see 'Materials and methods'). If pollinators foraged locally, then they would not have been exposed to flowers with a UV bullseye pattern. Further studies using naïve insects will help to determine whether preference for the bull's eye pattern is innate or learned for these insects.

Implications for Naturally Occurring Variation in UV Pattern

This study was aimed at understanding the function of UV pattern experimentally, but it has implications for species with naturally occurring intraspecific variation for UV pattern. For example, populations can be polymorphic for the presence or absence of the UV bullseye (Cruden 1972; Naruhashi & Ikeda 1999) or can display quantitative variation in its size (Koski & Ashman 2013), and this variation can be heritable (Yoshioka et al. 2005; Syafaruddin et al. 2006; Koski & Ashman 2013). If the pollination dynamics observed in the present study are representative of natural conditions in A. anserina and other systems, then individuals that lack the UV bullseye may attract fewer insects than those with UV pattern. However, this may not necessarily lead to significant differences in foraging rates (Fig. 3d), which correlate positively with female and male fitness (e.g. Galen 1989; Ashman 2000). This lack of difference could explain the maintenance of extensive variation for the size of the bullseye in many populations of A. anserina, including fully UV-absorbing (Koski & Ashman 2013). However, our results indicate that flowers with an aberrant bullseye (inverse) would experience reduced fitness (Fig. 3f), and interestingly, this phenotype is not known to exist in A. anserina, or, to our knowledge, any other species.

Variation for UV pattern exists in many species (e.g. Rieseberg & Schilling 1985; Naruhashi & Ikeda 1999; Koski & Ashman 2013) and our study shows that pollinators respond to this variation. However, understanding the direct fitness consequences of this variation requires further study. We suggest that factors other than pollinators should also be considered. For example, UV-absorbing compounds in petals that give rise to UV pattern could protect against abiotic stress (cold/heat. Rivero et al. 2001; UV radiation, Jansen, Gaba & Greenberg 1998) or florivore damage (Gronquist et al. 2001). Conflicting selection pressures of these types may maintain variation in pattern (e.g. Frey 2004). Selection analyses that utilize natural variation are much needed to deepen our understanding of the functional significance of UV floral patterns.

Acknowledgements

We thank the staff at Pymatuning Laboratory of Ecology, and the Linesville State Fish Hatchery for access to field sites, N. Morehouse for use of the reflectance spectrometer, T. Kim, T. Byers, and J. Golden for field and lab assistance, the Ashman Lab, UPitt's E & E faculty and graduate students for discussion, and Diane Campbell and two anonymous reviewers for insightful comments on the manuscript, and PLE McKinley and Pape awards, an SSE Rosemary Grant Award, a BSA J.S. Karling Award, a Sigma Xi Grant-in-Aid-of-Research, and a NSF Graduate Research Fellowship to MHK, and NSF (DEB 1020523) to TLA for funding.

Data accessibility

Data deposited in the Dryad repository: http://doi.org/10.5061/dryad.bm558 (Koski & Ashman 2014).

Ancillary